An Ensemble Classifier for Predicting Eukaryotic Protein Subcellular Locations

摘要

Eukaryotic protein subcellular localization is an important and challenging problem in cell biology and proteomics. To tackle this problem, eukaryotic protein sequences were represented as amino acid composition and gapped pair amino acid composition, with and without 9-letter exchange.Based on such a representation frame, an ensemble classifier was developed by fusing ten basic individual K-local Hyperplane Distance Nearest Neighbor (HKNN) classifiers through majority voting scheme. Experimental results obtained through 5-fold cross-validation test on the same protein dataset, which contains eukaryotic proteins among 12 locations, showed a significant improvement in prediction accuracy over existing methods.